Вы находитесь на странице: 1из 19

AXK.

~LS

OF

PHYSICS:

16: 158-176

(1961)

Generalized

Free

Fields
0.

and
\;li.

Models
GREESBERGt$

of Local

Field

Theory*

Physics

Department Institute

and Laboratory oj Technology,

for Nuclear Science, Cambridge, Massachuseffs

dlasrach~~sefts

C;eneralized free fields are mutator is a c-number hut equation. Some properties of appropriate generators of the tor, spectrum, locality, and are used to construct a family eyuivalent to any (generalized)

a generalization of the free field in which the comdoes not satisfy a homogeneous Klein-Gordon generalized free fields are examined including the inhomogeneous Lorentz group, the number operaasymptotic fields. These generalized free fields of local field theory models which is not unitarily free field. I. INTRC)I)UCTIOiX

Local quant,um field theory is thr main theoretical tool in the study of elementary particle physics. However, up to now, there is no example, which is known to he consistent, of a local quant,um field theory in which wattrring or react,ions occur. Seit,her is there a theorem which states the existence of such theories. liurt~hcrmorr, there are few examples of any kind which ratisfy the reyuirement,s of local field theory. In this paper WC introduce generalized frrr fields, &nd use t,hese fields to construct a set of models which satisfy the rcquircments of local field theory, hut, in which no scattering or reactions occur. k?rst we drfinc these generalized free fields and rsnmillr romc of their propertics, including their generators of translat8ions and Lorcntz transformations, number operator, spectnlm, locality, and asymptotic fields. WC then uw thcsc generaliztd free fields to con&xct a set of field throry models which satisfy the requirements of local field theory and are not (unitarily) cquivnlcnt to auy (generalized) free field. We t)ake as the requirements of local field theory the following four properties: I---relativistic transformation properties, II-unique, uormalizable, invariant vacuum state \ko , and no nrgativr energy states OI states of space-like momenta, III-local commutation relations, Iv-complctencss of the set of states obtjained by applying polynomials in the smeared field
* This work is supported in part through U. S. Atomic Energy Commission (ontract. At(301-1)2098, hy funds provided by t,he U. S. iltomic Energy Commission, the Office of Naval Research, and the Air Force Office of Scientific Research. t National Science Foundation Post-doctoral Fellow. 1: Now at Physics Ikpartment, University of Maryland, College Pz,rk, Maryland. 158

(~ESER.~I,IZICI)

FREE

FIELIS

159

operat*ors to the vacuum state \EO. Our models are characterized by an arbitrary measure and an arbikary symmetric function of two variables. Sonc of there models lead to scattering or reactions; they all have N-matrix equal to onr; honever the vacuum expectation values and Greens functions of there models differ from t)hose of the free field theory. One reason that S = 1 is, roughly speaking, that, the vacuum expectation values for these models, considered in momentum space, agreewit,h the vacuum expectation values of the free field in the neighhorhood of t,hc stable particle mass shell p* = We, even though the vacuum cspertation values of these models differ from those of the free field away from this mass shell. Xone of these models have canonical commutat,ion relations. In Section 2 we state properties I through IT, define the generalized free fields, and study their properties. Section 3 discusses the set of models of loral field t,heory constructed using generalized free fields. The last section sunnnarize:: t,hc work of this paper.
II. REQUIREMENTS PROPERTIES OF LOCAL FIELI) OF (:ESERALI%E;I) THEORY; FREE I)EFINITION FIEI,l)S ASD

A. REQUIHEMENTA

OF LOCAL FIELD THEORY'

Property I requires that the field A (s ) transform mider a Gtary representation of tbe inhomogeneous Lorentz group, I( a, A), where a is a space-time translaGon, and h t L+r (the orthochronous group of Lorentz tra,nsformations of det,erminant# one), as a scalar field

C(a, A),4(x)U(a,
Property II requires that

A-

= A(Ax + a).

(1)

CT(u, ii)*0 or I%0 = 0, and

= \Il,

(gal

dd\ko = 0,

(2b)

where 90 is the unique, normalizahle, invariant vacuum st,at,e of the theory, and I and JP are the generators of spare-&me translations (i.e., the momentumenergy operat)ors) and generat)ors of homogeneous Lorentz transformations (i.e., the moment, of energy and angular momentum operators) of the theory respectively. The spectrum of P0 must be bounded from below by zero. Property III requires that the commutator of the field operators at two different points vanish for space-like separation of the points, L4(2), A(!/)1 = 0, (.r - y)' < 0. 1.3)

(Note that we use the metric for which s = so2 - x2. ) Propert,y IV requires that
1 For an extensive discussion of the requirements of local field theory, see Ref. 1

160 any state \k can be approximated 5


j=U

GHEENBERG

in the mran by a state of t,he form . . . rj)A(sl) . . . A(r,)qo

/ dr, . . . d4.#(x1,

for N finite. Finally, we introduce t,he notions of canonical fields and asymptot,ic fields (1-s). Property V- the canonical property-requires that (a) there are operators A (x, t), T(X, t j which satisfy the equal time commutation relations

M(x, t), dY, Gl = iS(x - Y), M(x, t), A(Y, t)l = Idx, t), T(Y, Gl = 0,
and (b) this pair of operators at a given time suffices t#odefine the whole theory in the sense of Property IV, where A(x, t) and a( X, tj replace A(s) and the integrat#ions run over d3x. Property VI-the asymptotic condit,ion-requires that out (a) the asymptotic fields Ai (2) of stable mass WLdefined by
0t

Ai (.r)

= lim

c-3 (where B $ convergent;


C = B(aC/&r

r-*CC
) - (aB/&r )C

C-U

exist in t#he sense of weak operator > of the smeared operators (that is integrate both sides of Eq. (4) wit,h f(x)dx, where .f is an arbitrary testring functiou in the Schwartz space s), out satisfy the free Klein-Gordon equation of mass 1~2,(0 + o?)A = 0, transform as scalar fields under ((a, A), have free field commutaGon relations, out out [Ain (x), A (yj] = iA,z(.c - y), and have a posit#ivefrequency part, which annihilates the vacuum,
2 We define positive and negative frequency and spacelike parts of a field A (2) by

A()(L)
where respectively.

= 11/(27rPl

s
e,(k)

flJkej(k)e-~-d(l/)

CP?/,

For a vector

j = zt k.

OP

S,

and = 1, 0,

= ti(z!zk)

or

e(-P),

e(k) and for a number

x.2 2 0, otherwise

ko > 0,

GEh-ERALIZEU
Out

FREE

FIELDS

161

A in +(.r)\k

= 0;

(1)) the asymptotic fields are complete in the sense that Propert,y IV holds with with iloUt or ilin (s) replacing i4(.r). B.
DEFKITIOK OF GEXEHALIZED IHEE FIELDS

We define generalized commutation relation4 [4(x), where A,z(s) 4(y)]

free fields

to be field operators3

which

satisfy

t,he

= iA(x

- y) = i lrn dap(u)A,z(.r

- y),

(5)

= --[i/(&r)]

[ d%e-~(k)6(k*

- a)

is the free field commutator function for mass a, and ~(a) is a positive measure5 of not too fast increase. We call p the Lehmann weight (5) of the field 4. In order that the generator of time kanslations, PO, have a positive definite spectrum (Propert,y II), it, is necessary that f#J(+yr)\k = c#F(a)\ko = 0, (6)

where Pa is the vacuum stat,e. From Eq. (ci j it follows that [+(x), 4(y)] = 0. This fact together with Eq. (6) implies that all matrix elements (x, @q) vanish, where x and Q are any st,ates in the Hilbert space belonging t,o 4. Thus +((x) = 0. As a concret,e example of a generalized free field to which we will refer later, we use the case A where there is a single disc&e mass II) in addition t,o a continuum, and p(a2) = s(u - 777) + fJ(u2); fJ(u2) = 0, a < (2m); du) > 0, CL22 (h)*. into its discrete and

For t#his example A we define a separation of the field 4(s) cont~inuous mass parts, &(.r) and +1(a), respectively:

3 For simplicity we assume that all the fields we discuss are neutral, scalar fields. 4 Wellner (4) has used similar commutation relations in a model of decaying elementary particles. Licht and Toll have shown that a complete, local field 4(s) which has a translation invariant commutator is a generalized free field. 5 In the present paper we assume that the space of state vectors is a Hilhert space, i.e., a space with a positive definite metric. If this assumption is not made. then p need not he a positive measure.

162 where

GREESBERG

&J(X) =s

d4ke- ik(x--u)0,2( i?)cjb( y) d41J, From

and O,z(li) is one in a small neighborhood of /? = TI? and zero elsewhere. bhe definitions of +0 and +l , we find the commutation relations

[ddt),~d~)l =
and

iAmz(:r - Y), [~J~(x:), h(y)1

= i/mI+edaa($)Aa2(z

- Y),

(7a)

bo(x.), &(!/)I together with &lyzNfo

= 0,

Vb)

= 0. equivalent equation,

(8) to a free field

Equations (7) and (8) require that 40(z) be unitarily of mass m so that +o(z) satisfies t,he free Klein-Gordon (0 + m%o(5) = 0.

If ~(a*) = 0, we recover the usual free field of mass 1)). We conclude t#his subse&on by showing that Eqs. (5) and (6) determine the t,heory of the single field Q(X) completely, provided we assume that $J(x) is the only field present in the theory. Our demonstration that, Eqs. (5) and (6) determine the t,heory rests on the fact that tzhe knowledge of all the vaccum expectat,ion values of the t)heory, {F(Xl where pP)(.J.l ) . . xx) = c*o, 4421) . . . d.rvNo), determines the t,heory completely (1). With no loss of generality, we assume t,hat F{ki = 0. Using Eq. (5) and (6), the Hermiticity of +(x) and the definition of +*(x), we compute F(x1 , x.2) = ($(x~)\Eo , $(xz)Po) = (PO, ++(.~,)4O+h)
x2) = i m d&(i)A:t(.rl /0 - x2),

. . xn);

I1 > O),

= (qo, [$+(.I&

c$(x2)]\ko)

= iA+(.r,

6 This assumption involves no loss of generality because if Fll(s) = (UO , +(r)Po) # 0, since by translation invariance then in place of @J(Z) we can consider d(s) = 4(x) - F(, F(l) is independent of s. Equations (5) and (6) still hold for C#J(CZ); but now fit!)(s) = (qO , c$(z)Y~) = 0. If we wish we can i-ecover the original 6 by adding F( to 6 later.

GESEH.ILIZED

FREE FIELDS

1G3

where
A:+'(X) = --[il(2~)"] 1 d'kC'k"@(1<)6(k' n").

In a similar may we can reduce any F to Fin- : F(r1 ( 22 , . . . .r,,j = (\ko ) c#(+(al)+(x?) . . . f#J(x*)*)

=ciA+)(.r~ ,x3,.TimI,..r,). - .riVrn-P(xg,. .. s;+l


Since WPknow F and F, by recursion we know all II (in terms of t,he give11 p) and thus we have determined the theory completely. C. RELATIVISTIC
~VARI~NCE; ~,OHENTZ TK~KSFORMATIOSS GEWXMT~RS 0~ TIUSSLATIOKS ~ZK;D

In this subsection we construct a rcprcsrntat)ion of the inhomogcnrous Lorentx group, [:(a, A), which transforms a gencrslized free field as a scalar Hrisenberg field, Eq. ( 1) ; with this representation Property I is Fatisfied for generalized frer fields. We consbwct r7(n, A) in t,erms of its Hcrmitian generators P, and JJ,,, , aiiglilar momentum tewor, the energy-momentum wctor and relativistit respectively. First we consider the translations, I(n, 1 ) = PYP [~/.a]. Equation (1 ) implies dl$,dr = i[f, , #J(x)], which leads us to guess that the I, are hilincnr function& of +( .r). With this guess, a straightforward complltntion lcnda to the rsprrssions (for thr cxse A wit,h one discrete mass)

164
and7 where

GREENBERG

(9d)

K(y) = 4/2@7$1 I,, b2/&)1&~(~).

(se>

Note that the PF' and P:" are four-dimensional nonlocal functionals of 4(x) local functionals of 40(r) ) in con(even though the P,? are three-dimensional trast to the three-dimensional local form of the usual free field translation operators8 Next we consider Lorentz transformations, U(0, A) = exp [(i/2)M,,y~Y]. Now Eq. (1) implies

Assuming

that M,, is a bilinear functional

of +(z),

we find the expression (lOa)

M,, = ML:' + Mb:',


where again M,,I is the free field functional of 40 ,

Al;;' = -

x) s&LO d3X[Z,CP$yrCJ, - X" 6:u'(x0, x)],


defined for frequency (creation)

(lob)
generalized (annihilaoperators free as an

7 The double dots in Eq. (Yd) stand for normal ordering which is free fields in the same way that it is for free fields; that is all posit,ive tion) operators are to be moved to the right and all negative frequency are to be moved to t,he left. 8 There is another way to write the continuum mass contribution field Hamiltonian (i.e., generator of time translations) in which integral over masses of free field Hamiltonians of a given mass. We

to the generalized PA is expressed write

u-here

is defined only for a* in the support of (r. Note that the operation which defines +.2(r) is a IS by 4(x) at a given t,ime nonlocal four-dimensional procedure so that PA . not determined even though it is given in t,erms of the set ( +oz(s), a2 in the support of g } at a single value of for the other Pk and for the MOP ~0. Expressions analogous to that for P, can be written and N(r).

GENER-4LIZED

FREE

FIELDS

165

and

iu:; = - s d4.r[a, - s,s&y sl(:I) (x)1.

(1Oc)

?;ote that the point, (0, 0, 0,O) has been cho,sen as the center of rotation (Lorentz transformation) for the operators M,,, . Equations (9) and (10) for P, and M,, define the representation of t,he inhomogeneous Lorentz group which is appropriate for generalized free fields; with this representation Propertly I is satisfied. The number operator for generalized free fields is N = N 4 Nil, where

= (i/2)

1 d3X [ml(x) of ~$0 and ,

$b:+)(J)]

is the usual free field functional y(1) where A::(x) The Appendix D.
SPECTRUM

1 da2 Ai% - y) d4.r d4y : 41(s) c#a, Cl/):, 2 C27r)2s c7(a2)

= [1/(27r)3]

/ d4k6(k2 - $)e-? for P, , M,, , and N.

gives momentunl
ASD LOCALITY

space expressions

Property II, that the theory has a vacuunl state and a positive energy spect,rum follows from t,hc construct,ion of P, , M,, which satisfy Eq. (2b) by virtue of Eq. (61, and the fact that any functional of generalized free field operators has a normal ordered expansion

The st)ate O\ko has contributions only from the crcaGon (negative frequency) parts of 4; thus the spectrum of PO is bounded below by zero. Property III, locality, holds because of the local property of A,z(.r - ~1) in the definition of 4, Eq. (5). E.
POSITIVE I~EFIKITEXEM COXDITIOAX (1)

In t,his subsection we show that the positivity of the Lehmann weight, p is a necessary and sufficient condition that the states created by smeared polynomicals in the generalized free field + acting on the xTacuum state have posit#ive

166

(:ItEENUEltG

norm, i.e., that t,his vector space be a Hilhert space. Necessity is immediate because it is already known (5, 1) that, the positivity of p is necessary for the positivity of /j +(.f)90 11 = (\ko , +(.&(f)\ko) where 4(f) = jd4.r+(rjp(s), and ,f is a testing function in the Schwartz space S. We show sufficiency direct)ly by computing the norm of any vector and finding that) it, is positive. ,\ny vector \k can be written \k = cz=O \Ir,, , where 9, = c j rli,l.zi,= crnimlr~ u n,J-:2 n P(.fT, a )90

is a vect#or in the n-yuant,um subspace, (,n,,\ are the occupation numbers for quanta in t,he st,ates (iJ and CIN{ni,] is the probability amplitude that the state with occupation numbers [ni,j is in t.he vector q,, . The wave functions f; are chosen to be a complete orthonormal set with the scalar product

Then the commutat)ion festly positive norm

relations

of 4(x),

Eys

(5) and (6), lead to the mani-

This completes the proof t,hat# a positive Lehmann weight in a necessary and sufficient, condition that the positive definiteness condit,ions are satisfied for the theory of the field 4; and also concludes our demonstration that Properties I through IV are sat,isfied for a generalized free field 4.
I?. hYMPTOTIC b.IELI)S out

In this subsection we study the asymptotic fields +in associated with 4 from the point of \-iew of Property VI. We find that) thesc asympt,otic fields exist and have t,hc properties which we collectively call Property VI (a). In particular r#loUt(.r) = p(s) = c#Yl(J), out. which relates +in by $o(r) = S--$p(r)S

and the S-matrix,

is t,he identity operator. Thus neither scattering nor reactions are described by a generalized free field. The st#atements of this paragraph follow straightforwardly by embedding the definition of the asymptotic fields, Eq. (4), in an arbitrary vacuum expectation value and performing the limit 7 ---) f = on the distributions t,hus obtained. DThe limit, in
the

sense

of

tlistrihution

theory,

GENERrZLIZED

FREE

FIELDS

167

However Property VI (b) is not satisfied; that is the asymptotic fields are not a complete set of fields for generalized free fields. Property VI (b) is equivalent to the existence of an expansion of 4(r) in terms of normal ordered products of @*l(r) = +0(x). If the expansion
($(s) = g1 SC&, . . . &,f(x gl, . . . .r y,,) :p(yl) . . . @(grt): (11)

exists, then t,he expansion coefficients

can be computed

in the following

way: (12a)

=s

d4y, d4y,f(21(x

-. gl, .r - ti)[iArnz(~l + ;A,z(y,

- zl)iA,z(u.~ - zs)iA,n(y,

- 2~) (12b) - x,,]

and in a similar way the vacuum expectation value of the lath interated commutator leads to an equation for f. On the other hand, the expressions on t)he left of Eq. (12) can be computed direct,ly from t#he commutation relation, Eq. (a), which yields the result

(b#d~,,@(~~)l)o &z(.r - 21) =


([. . . b(J), Comparing 4Ji(G)l, . . 4in(z,l)])0 = 0, n 2 2. Eqs. ( 12) and ( 13), we find t#hat .fw .f(y1 sincef(kl, . . . I;,) Fourier transform.) is clearly wrong. We generalized free field = Y!/), n L 2,

(1Sa) (131~)

(143) (14b)

) j/2 ) . . . yn) = 0,

has support kc = n?. (Throughout this paper a tilde means Thus the result, of our expansion is 4(x) = $+(x) which conclude that the asymptotic fields are not complete for a theory.lOm
(1) = 0, otherwise ,,p

lim i+*oc

allows a straightforward discussion of the asymptotic limit for generalized free fields. I The fact that the LSZ asymptotic fields fail to be complete in this simple case leads to the question: Are these asymptotic fields complete for a local field t,heory with interaction? The assumption that the answer is yes seems tjo be part of the orthodox position in local field theory (2); however, this completeness assumption has not been exploited in many applications of local field theory.

168
III. CLASS OF MODELS

GREENBERG

OF

LOCAL

FIELD

THEORY

Now we use the generalized free fields discussed above to construct models of local field theory. In particular we construct1 models which can be exhibited in closed form and are exactly local. The problem of finding Heisenberg fields which are local in successiveorders of some approximation scheme is also interesting; however we do not discussthat problem here. We study a Heisenberg field A (x) which can be expandedI in a series of normal ordered products of a generalized free field 4(x),
A(x) = 4(x) + g2 1 d4yi . . . d4y,, (15) .f(x - y1 , . . . x - yr,):c$(.rl) ... $(yn):

where f((vl , . . . yn) = f( Ayl , . . . Ay,) is a real Lorentz invariant function which is symmetric in it)s arguments. We assert t)hat Property III, locality, is the only one of the four requirements of field theory that remains to be imposed on this expression for A(x). Property I, relativistic invariance, is satisfied with U(a, A) equal to the operator determined for the generalized free field @B(Z) in Subsection II (C). This fact follows from
U(a,A)A(x)U(a, = gl A)- - y,, + a . . . .I: - yn) :+(Ayl + a) ... +(Ay, + a):

j d4y1 . . . d4y,rf(.r

= 2 j d4X1 . . . dx,,f(hx u=l

x1 , . . . hx i- a - z,,) :$(G)

. . . $(z,) : = rl(Ar

+ a),

I1 Note added in proof: Haag and Schroer (to be published) have proved that a generalized free field whose Lehmann weight contains a continuous contribution is not complete in a time slice. 12 An analogous situation holds with respect to the canonical property. Property V(a) is satisfied with 4(x, t) and 4(x, 1) as canonical fields and a nave-function renormalization constant in the relation [6(x, tJ,t#i(y, t,] = i s0 However Property V(b), completeness of this pair of operat,ors at a single time, does not hold unless the field is an ordinary mass VL free field. We do not exclude the possibility that some other pair of canonical fields at a single time is complete. Single time completeness is one of the assumptions of Haags theorem (6) ; therefore there is no straightforward extension of Haags theorem to fields related unitarily at a given time to a generalized free field. I3 Haag (6) has considered the expansion of a Heisenberg field in terms of a mass 7~ free field. Sudarshan and Bardakci (to be published) have used the Haag mass 17~free field espansion to construct local models analogous to t)hose considered in this section. co da* p(aM(x y).

GENERALIZED

FREE

FIELDS

169

where we have used the change of integration variables .zi = ~yi + a, the Lorentz invariance of the fCn, and-to include the term $(.r) in the sum-have introduced f(r) = 6(s). Property II, positive energy spectrum, follows from t)he discussion in Section II (D). The positive definiteness conditions follow from a discussion similar to that of Section II(E). Thus only locality requires a det,ailed discussion. To discuss localit,y we observe that, the commutator [i4 (x), A(y)] can be romput#ed explicitly in terms of normal ordered products of 4(x). The result has the form

.I

d4X1 . . . d4z,

g (

s -

.le + y;

.r + x1 ) . . . 2

2, >

:dzd

..

44X,8)

:,

whereg(z; symmet,ric

x1 , . . . x,,) = y(hr; & , . . . k) is real, Lorentz invariant, and in ( z1 , . . . 2,). The locality of A (2) is equivalent to the requirement gyx; 21 ) . . . x,,) = 0, x2 < 0, for all xi , and all n. 2 0.

Since the gCn are funrt,ionals of the f and t,he Lehmann weight p, t,he locality of 9 imposes restrictions on the jand p. We now rest.rict ourselves to the simplest nontrivial case of Eq. (15) in which only the linear and bilinear terms in 4 are retained, and find sufficient, conditions for localit,y in this case. We discuss the locality of [il(.r), L4(y)] = / d41i1d4k2 . exp -i(kI + .r + y . 2 - i(k, - k,) . y

li,)

[X(lcl),

B(k,)]

in terms of the commutjator of the momentum space fields, and rall t,his momentum space commutatjor loral if the corresponding x:-spacae commutator is local. Note that, it, is the dependence of [A( k,), d(k,)] on p = li, - kc, which ip relevant for locality. The moment~um space field is given by

J(k) = l+(k)+ [ dpl(; + 23,; :,(;+,>,(; -23)

-9):

(16)

where a numerical fact or has been absorbed in f. The commutator [A( kl), A( ,+?)I has t,hree terms cont,aining normal ordered products of zero, one and t,wo + operators. The term wit,h no 4 operators is necessarily local, since it is the vacuum expectation value of the commutator (6). The terms with one or two + operators are 2[t(lil)p(X:j)J( -kl , kl +li?) - t(li?)p(X:>').T( -kXr , k1 + X2)] r$(lil + kz) (17a)

170

GREENBERG

respectively. To discuss the locality of these terms we use the theorem t,hat a necessary and sufficient condition for a function 6(q) to have a Fourier transform which vanishes for x2 < 0 is that, i(q) be representable in the form (7) I;(q) = s d4U s ds c(qO - .u)d[(y - uy -s]\k(u, s).

[If, in addition, /i(q) vanishes in a given region bounded by t,wo spacelike surfaces c1 and g2 , then \k(u, s) vanishes outside a region which depends on (or and (~2. We do not have t,o insert explicitly the analog of t,his support condition for our commutator [A( aI), B( k2)] ; the support condition for matrix elements of [A(h), &WI is guaranteed since we st,art# from an operat,or expression for A(lz) .] We now exhibit the most general local expressions with one or two 4 operators using t.he above t#heorem: for one + operator,

and for two 4 operators,

(Mb)
Ii2 :. )

.- ( .6

h*kpp >

4 (

kl + -2-p

If we can find expressions for 1, and for II and \k2 so t,hat Eqs. ( 17a, b) have the same form as Eys. (18a, b), respect#ively, then our held A (x) will he local The ansatzes

C;ESER.~LIZED

FREE

FIELW

171

change both expressions

with

one + operator

to

and change both expressions

with two + operators

to

Therefore

is a suflicient condition field of the form A(x) = 4(e)

for t,he locality

of A (x). In r-space our result, is that, any

+ 1 r((.c

- yd?,(.x - yz)?) :+(y&(yd

: d4yld4ya

(19,

is loc+al. For the case where 4(z) field .?jk) reduces to


X(/i)

is chosen to he a free field, 4(r),

of mass

nz,

the local

= &(k)d(k

- WL) + y(??Z, VI) /d4dk(;+,)

.q;+q

- wl$&(;-+[(g -q - $1 :,

C:c do not know if there is a mno;iical field a( x, t) m-hic*h, together with 4 (x, t 1, satisfies Iroperty 1. However, W can auswcr the cluestim: Is A(x, t) the cnnoni~nl field T-(X, t )? A%\s might espivt for a theory which differs so we tnur*h fro:11 a c~onvc~~tionnl uonderivati rc mipling theory, the nuswer is m. \\r show t hi+ hy finding the co!iditioiis that [;l(x,
t),

A(y, f,J = ciS(x - yl.

172 These conditions, which

GREENBERG

follow from a straightforward 10 p(s)y(s, s) ds = 0, all s0

calculation, s

are (20a)

and * s0 p(s)y(s, s)y(s, s) ds = 0, all s, s. (20b)

If we set s = a in Eq. (20b), we find m p(s)[y(s, s)] s0 Then the reality and symmetry

= 0,

all

s. of p require

of y, and the positivity = 0, all s, sn,

y (s, s)

so that our field A(z) reduces to the generalized free field 4(r). Now we discuss the scattering matrix for these model? from the point of view of the LSZ formalism (2, 3, 1) and show that S = 1, i.e., no scattering or reactions occur. Since the generalized free field does not produce scattering or reactions, we can conclude that S = 1 if we can show that only the first term on the right-hand side of Eq. (16) or Eq. (19) contributes to the asymptotic fields. Now the term in A(z) which contains :~(YJ~)c$(Y~): contributes only to matrix elements (9, :+(yl)+(yl) : x) in which t(he numbers of 4 quanta in \k and x differ by zero or two. Thus t,o determine the comribution to the asymptotic limit from this term it suffices tjo consider the asymptotic limit applied to ,4 (x) in the matrix elementsM2 = ($*(pl)\ko ,A(.r)&*(pz)\ko) and IIf1 = ($*(P~)c$*(P~)\~o, -4(x)*0). We now compute the asymptotic fields in these two matrix elements in momentum space using the momentum space equivalent of Eq. (4))
out

(12) =

lim e(k)G(k t+*%-

- m) / dqX(q,

k) (k + ~~)e-~~*~-~~,

as the definition

of tfhe in and out fields. We find

(21) . lim dq,y(p;,(p: t+*Z? s + CJ)~ - (pl + k))(k + ~)~-ino+k2+m2r~.

14 It can he shown quite generally that any model which is constructed in terms of a finite degree normal ordered expansion in (generalized) free fields either has S = 1, or has a nonunitary S-matrix.

GENERALIZED

FREE

FIELDS

173

Riemann-Lebesque lemma this limit vanishes unless as a singularity at which q + (k + v?)~ = 0. Apt, (PI + q12 - (PI + W21 h However, a function y depending on the variables indicated cannot have such a singularity, and thus the limits in Eq. (21) vanish. A similar argument shows that the asymptotic limit applied to t.he matrix element M2 also vanishes. We conclude that out out A (32) = fp (z) = c#q(2-) and X = 1. The heuristic reason for this result is simple. To get AoUt # A(k) we must have principal value singularit,ies in matrix elements of A(k) at k2 = w?. Equation (16) for A(k) contains the c-number functionf( (k/2) + p, (k/2) - p) which a priori might depend on the three scalar products ((k/2) + p), ((k/2) - p), and k2. Our condition for locality is that there be no dependence on this third variable, li. Thus locality conspires to eliminate from our control just that variable which is relevant for the difference between AoUt and A. In this paragraph we show that the field A (2) which we have constructed is in hhe Borchers class (8) of both its generalized free field 4(z) and t#he mass m part, &(.r), of this generalized free field. To do this, we compute the commutator [A(k,), ~$(k?)] and show that it is locIal. The calculat,ion is

According

to

the

where we have introduccd the variables p = ,& + k, second term is in the local Jest-Lehmann-Dyson form 0, (X - g) < 0 and il is in the Borchers class of 4. ,4 the same result for the commutator [A(lz,), $o(l~s)6(ksZ

and (1 = k1 - k2 , and t,he (7). Thus [A (z), 4(y)] = similar comput8at,ion yields - UZ)].

We have studied a simple generalization of the free field, in which t)hc commutator is a c-number but can have t)he space-timc dependence of the vacuum expectat)ion value of a commutator of arbitrary Heisenberg fields, rather than satisfying a homogeneous KleinGordon equat8ion. For the example A in which thc generalized free field is taken to have a discrete mass part 40 at mass 1~1, and a continuous mass part, +1 starting at some higher mass, the generators of trxnslations and Lorentz transformations all have a term bilinear in 4 which has a nonlocal four dimensional kernel. The asymptotic fields exist and are given by

174

CREENBERC T

4in (x) = 40(x); however, these asymptotic fields are not, complete. Similarly 4 and 4 at a given t#ime satisfy t,he canonical c*ommut8ation relations, hut, are not complete. Csing these generalized free fields we constructed a set of models of field theory in which the Heisenherg field sat,isfies the requircment,s of rclat,ivist,ic covariancc, spectrum, localit#y, and part of the asymptotic condit,ion. The asymptotic fields for these models are not, complete. Xo equation of mot,ion or time slice condition is known for t,hese models. These models have S = 1, and are in the Borchers class of their generalized free field. We conclude by mentioning an idea, not yet developed, which is suggested by generalized free fields. Generalized free fields + can he considered to he the lowest order case of models of fields whose commutat,or can be expanded in products of Heisenberg fields cont,aining up to a given number of factors. For generalized free fieldstherelations[$(r), +(yj] = iA(r - y) and4+(r)% = +(x)@ = Odetermine the field. Similarly, for the next more complicated case, the commutator

[A(r),A(y)l

= iA(s

- y) + [ d4zJ(x

y,$J

2) A(x)

(22)

and the requirement,s il+(r)k,, = .4(X)QS = 0 ( 23 )

determine all the vacuum expectation values by recursion and t,hus again determine the field. Using argumentIs similar to t,hose discussedabove, it, is straightforward t)o choose.f so t,hat the commutator of Eq. (22) satisfies translat,iou invariance, 1,orentz invariance, and localit,y. One can also see that Eq. (23), toget,her wit,h translation invariance, insures a positrive energy spectrum. However it is difficult t,o find conditions on p and j such that, (a) the commutator of Eq. (22) satisfies the Jacobi ident,ity, and (b) the vacuum expectation values generat,ed by recursion from Eqs. (22) and (2X satisfy the positive dtfinit)enrss conditions (1). Models satisfying Eqs. (22) and (23) have a considerably more complicat,cd structureI than those considered in Section 3 ; in particular t,hcy do not, have a finite degree normal ordered expansion in terms of a gcncralisecl free field, and thlts may give rise t,o scaatteringand reactions.
APPlOYI)IS: MOMI~XTUM SPACE P:SPRESSIOiY;S FOR I, , iIf,,, ANI, 1V

Define t,hr Fourier transformed ( momentum space)field $( I) by

I5 Johnson has remarked that the mmmut:~tion generators of a Lie group of ir1finit.e order.

relations

of IlGi.

(22) are

those

of the

GENERALIZED

FREE

FIELDS

175

Corresponding

to the separation 4(q)

@(.I+) = &I(X) + $I( x), write + 41(q). are a4p + q),

= io(qh%22 - )i) in momentum space

The commutation [c&(p)6(p d),

relations

&(q)6(qZ [&(p)6(pJ

n?)] = [l/(2a>lt(p)6(p2 ))12), &(q)l = 0, + 9).

k&(p), The generators pi) and

h(9)]

= ~~/(2~)l~(P)~(P~~(P

of translations =

are P, = Prco + P,?, where

K%r)/m [ d4pe(p):a -p)p, &(p)6(p2 - m) :

I J,:l) = [(2a)/2] The generators ML;) = -y and J/f;; = -- (27r)i 2 s Finally, the number N and of Lorentz

d4p14p)l&J2N :A( -P)P, 41(P):.


are M,, = AI,? + *lrb:, where aBo(p) p, dp s(p2 - n?):

transformations

j d4pc(p) :$0(-p)

aMp) P,----w

operator is N = N(O) + N(l), where = [(27r)/21/ d4p :&o( -pMo(pM(p - m):

A+) = [(27r)3/21 [ dp[l/a(pN


RECEIVED:

:d;1(-p)&(p):.

June 23, 1961

We thank Professors P. G. Federbush, It. Haag, F. E. Low, and E. C. G. Sudarshan for critical discussions of this work. We also thank the National Science Foundation for a postdoctoral fellowship during the tenure of which this work was done, and Professor N. H. Frank for the hospitality of the M.I.T. Physics I>epart,ment during the past two years.

176

GREENBERG

REFERENCES 1. A. 2. PI. S. WIGHTMAN, Phys. Rev. 101, 860 (1956); and ProhlPmes mathematiques de la theorie quantique des champs, University of Paris lecture notes (1957). H. LEHMANN, K. SYMANZIK AND W. ZIMMERMABN, Nuovo cimento [lo], 1, 205 (1955). We refer to this paper as LSZ. 0. W. GREENBERG, Thesis, Princeton University, 1956 (unpublished). M. WELLNER, 1959 (unpublished); A. L. LICHTAND J. S. TOLL, Nuovo cimento [lo], 21, 346 (1961). H. LEHMANN, Nuovo cimento [9], 11, 342 (1954). R. HAAG, Kgl. Danske Videnskab. Selskab, Mat.-fys. Medd. 24, No. 12 (1955); I>. W. HALL AND A. S. WIGHTMAN, Kgl. Danske Videnskab. Selskab, Mat.-fys. Medd. 31, No. 5(1957) ; 0. W. GREENBERG, Phys. Rev. 116, 706 (1959); R. JOST, 1959 (unpublished). F. J. DYSON, Phys. Rev. 110, 1460 (1958); R. JOST AND H. LEHMANN, Nvovo cimento [lo], 6, 1598 (1957). H. J. BORCHERS, Nuovo cimento [lo], 16, 784 (1960).

4.
5. 6.

7. 8.

Вам также может понравиться